Re: ap24-py310-python-3.5.0nb3 PLIST issue

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Sun, Aug 13, 2023 at 02:05:15PM -0400, Greg Troxel wrote:
> Joerg Sonnenberger <joerg%bec.de@localhost> writes:
> 
> > On Sun, Aug 13, 2023 at 03:29:22PM +0200, Martin Husemann wrote:
> >> I still consider this a serious pkgsrc bug - the build order of pkgs
> >> should not matter (even if this may by plain luck hide this issue in
> >> full bulk builds).
> >
> > It's more a setuptools bug really. The only way to fix it is to make
> > more things than necessary depend on setuptools. But given that
> > distutils is essentially dead, that train has passed.
> 
> That's not particularly constructive.  I agree with what I think you
> intend, that it's not really a pkgsrc bug, but rather pkgsrc's failure
> to work around upstream bugs.
> 
> But, regardless, it is wrong in pkgsrc if building a package obtains a
> different result depending on what was installed.
> 
> Are you really saying that you think it is completely intractable to
> patch this package so that it will build the same regardless of
> setuptools installed or not?

I'm saying that this kind of thing has happened over and over again over
the years and is a direct consequence of a setuptool "feature". It's
generally easy to patch up (e.g. just include egg.mk), but unless
someone actually hits it, it won't get magically detected.

Joerg